基于深度学习方法分类红团簇星与红巨星分支恒星  

Classification of Red Clump Stars and Red Giant Branch Stars Based on Deep Learning Method

作　　者：王冠宇 罗杨平 黎鑫 李启达 WANG Guan-yu;LUO Yang-ping;LI Xin;LI Qi-da(School of Physics and Astronomy,China West Normal University,Nanchong Sichuan 637009,China;Yunnan Observatories,Chinese Academy of Sciences,Kunming Yunnan 650216,China)

机构地区：[1]西华师范大学物理与天文学院,四川南充637009 [2]中国科学院云南天文台,昆明650216

出　　处：《西华师范大学学报（自然科学版）》2025年第1期86-94,共9页Journal of China West Normal University(Natural Sciences)

基　　金：国家自然科学基金项目(12173028)。

摘　　要：为了获取更纯净的红巨星样本,采用基于深度学习的方法对来自APOGEE-2的2万多颗星的演化类型进行了分类;使用APOKASC-2中的4216颗星为训练集,将神经网络训练出的模型应用到红巨星分支恒星和红团簇星的分类中;以APOKASC-2中的1807颗星为测试集,模型应用到测试集的结果显示,该方法的均方误差、均方根误差、平均绝对误差分别是4%、20%、9%。接着将模型应用到来自APOGEE-2的2万多颗星,得到结果的均方误差、均方根误差、平均绝对误差分别是5%、24%、8%。此外,分析使用了LAMOST DR7的数据进行测试,结果显示此方法的准确率目前能达到98%。通过对比,我们的方法与部分其他分类方法的准确率基本一致。未来可以将此方法应用于更大样本的红巨星分支恒星和红团簇星的分类工作。To obtain a purer sample of red giant branch stars,the study classify the evolutionary types of over 20000 stars from APOGEE-2 with deep learning method.The model is trained using 4216 stars from APOKASC-2 as the training set and applied to classify red giant branch and red clump stars.Using 1807 stars from APOKASC-2 as the test set,the result obtained shows that the mean square error,root mean square error,and mean absolute error of the method applied are 4%,20%,and 9%.Applying the model to more than 20000 stars from APOGEE-2,the mean square error,root mean square error,and mean absolute error of the results are 5%,24%,and 8%.Additionally,the analysis is tested with data from LAMOST DR7,demonstrating a current accuracy of 98%.The accuracy of the method in this study is basically equal to that of some other classification methods after comparison.Therefore,our method can be used in larger sky surveys for the classification of red clump and red giant branch stars.

关 键 词：红巨星支 机器学习 恒星参数 星震学 红团簇星 

分 类 号：P145.9[天文地球—天体物理]